We cannot exaggerate individual facial
components using the in-betweeners. We therefore define rules to
exaggerate the shape of a single facial component.These rules
mostly involve scaling and rotation with magnitudes determined by
comparing the component shapes of the given face with those of
average face.
Facial Shape
We divide the shape of the head into two
rectangles. The first rectangle includes the upper part,above the
eyes, and the second includes the lower part,below the eyes.These
rectangles are sub-divided into left and right parts by a vertical
line from the peak of the head. Figure 3(b-2) shows these rectangle
with dashed lines. The degree of haggardness or roundness of the
face can be estimated using the distance between the line
connecting from p 1 to p 2 (in Figure 3(b-2)) and the mean of all
the points in the left jaw region. The exaggeration parameters can
be extracted by comparing the degree of haggardness or roundness to
that of the average face (Figure 3(b-1)). Then the original feature
points (Figure 3(b-2)) are sheered to become exaggerated feature
points (Figure 3(b-3)) using the proportion specified by the
exaggeration parameters.